“…Additionally, Oh et al 6 incorporated user priors regarding moving objects into the low-rank model and improved the performance. Inspired by the successes of deep learning models in numerous vision tasks, [7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26] Yan et al 27 integrated spatial attention mechanisms into deep networks, which effectively mitigate misaligned content during HDR image reconstruction. However, motion removal-based methods, particularly in the presence of large-scale object motions in LDR images, tend to exclude a considerable number of pixels before merging the input LDR images.…”